GraphRAG — References

Retrieval that uses graph structure, the community detection it depends on, and the baseline it has to beat.

Graph-structured retrieval for generation

• From Local to Global: A Graph RAG Approach to Query-Focused Summarization — Darren Edge, Ha Trinh, Newman Cheng, Joshua Bradley, Alex Chao, Apurva Mody, Steven Truitt, Dasha Metropolitansky, Robert Osazuwa Ness and Jonathan Larson, arXiv preprint, 2024. Entity graph construction, community detection, hierarchical community summaries, and the query-focused summarisation task that motivates them. Cited by name.
• Microsoft GraphRAG — the reference implementation of that pipeline, and the most direct source for its build-time cost profile.
• Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (Lewis et al., 2020), arXiv:2005.11401 — the passage-retrieval formulation that any graph-based approach has to be measured against.

The structure being traversed

• Fast Unfolding of Communities in Large Networks — Vincent D. Blondel, Jean-Loup Guillaume, Renaud Lambiotte and Etienne Lefebvre, Journal of Statistical Mechanics, 2008. The Louvain method for detecting densely connected regions, the usual basis for choosing what to summarise. Cited by name.
• From Louvain to Leiden: Guaranteeing Well-Connected Communities — V. A. Traag, L. Waltman and N. J. van Eck, Scientific Reports, 2019. The correction to Louvain that avoids internally disconnected communities, which matters when a community is about to be summarised as one topic. Cited by name.

Reasoning over retrieved structure

• Think-on-Graph: Deep and Responsible Reasoning of Large Language Model on Knowledge Graph — Jiashuo Sun, Chengjin Xu, Lumingyuan Tang, Saizhuo Wang, Chen Lin, Yeyun Gong, Lionel M. Ni, Heung-Yeung Shum and Jian Guo, ICLR, 2024. Iterative, model-guided traversal of a knowledge graph as the retrieval step. Cited by name.
• Unifying Large Language Models and Knowledge Graphs: A Roadmap — Shirui Pan, Linhao Luo, Yufei Wang, Chen Chen, Jiapu Wang and Xindong Wu, IEEE Transactions on Knowledge and Data Engineering, 2024. A survey of the ways graphs and language models are combined, useful for placing a specific design among the alternatives. Cited by name.

No verified reference is recorded for incremental maintenance of community summaries, for degree capping during expansion, or for routing between graph and passage retrieval by question shape. These are open engineering problems rather than settled results.